SRD_V7 - Standard Reasoning (SRD) Model (V7)

Dataset

• Source: CoT_reasoning_unsloth.jsonl
• Examples: 9,340
• Format: messages[] chat format

Training Configuration

Parameter	Value
Learning Rate	0.00015
LoRA Rank	32
LoRA Alpha	64
LoRA Dropout	0.0
Target Modules	All (MLP + Attention)
Epochs	2
Batch Size (effective)	16
Warmup	3%
RSLoRA	Disabled

Training Results

• Training Time: 1.38 hours
• Final Loss: 1.2049

Part of Experiment

• kinzakhan1/CRD_V7
• kinzakhan1/SRD_V7 (this model)
• kinzakhan1/MIXED_V7